Group Event Early Warning Research Based On Big Data Of Network Petition

China is in a critical period of transformation,and social contradictions have become prominent.As a barometer of social problems,letters and visits are sensitive to social demands.After the network of China's petition work,all relevant information such as letters,visits,telephone calls and party secretary's mailbox are also connected to the network petition platform,which makes the data set of the network petition platform have the characteristics of multiple sources and heterogeneous,and concentrates on a large number of public opinions and conflict events.Social risk early warning is an important way to improve the country's ability to control risks.As a form of concentrated reflection of social problems,petition data can provide more effective information for social risk early warning.The main purpose of this study is to use the network petition data to predict the types of petition events that may develop into mass incidents,and to classify them into police levels,so as to provide a reference for the government to avoid the mass incidents that evolve from the petition.This article will network reporting events of Yunnan province from 2007 to 2018 as the research object,by using the method of content analysis and social semantic screening through the network of social risk early warning most predictive complaint reporting event types,and USES the analytic hierarchy process(AHP)for the early warning index system weight assignment,k-means clustering method is used to partition the police,to obtain early warning model;After using the variance analysis and verifying the early warning model of the group case in the petition case,the obtained early warning model has good validity.The results of this study show that: in the early warning model,there are two types of huge police cases and events,including housing demolition/land expropriation compensation and resettlement,salary and welfare;There are four types of serious police incidents,including disorderly conduct,health,and family planning management,repeated letters and visits,and social security.There were eight police incidents,including housing property transactions,environmental protection management,public security,urban and rural construction,village management,organization and personnel,inaction,and social assistance.There are three types of minor incidents,including management and charging,disaster management(natural,minor man-made)and market supervision.The early warning model corrected and determined by analysis of variance shows good validity.At the same time,the RBP neural network is used to analyze the time series of specific events,which makes a useful exploration for the trend of more precise warning of an event becoming a group event.The display method is scientific and effective,and the result is good.Finally,according to the Previous studies?Value-added theory and early warning model,the following Suggestions are put forward: 1.Do your polling ahead of time to make policy sense 2.Pay attention to ensuring that law enforcement and other administrative staffs contact the masses with a good working attitude;3.Pay close attention to online public opinions and public opinions at any time and make early warning;4.The government shall cooperate with relevant non-governmental organizations to make full use of the petition data and monitor social conditions sensitively;5.The government should pay attention to training and educating public opinion leaders to ensure the correct guidance of public opinion.